Research On Image Fusion Methods Based On Unconstrained Optimization EMD

With the expansion of the application range of multi-source image data,more and more demand for accurate and rapid acquisition of multi-view information through image fusion is proposed.Empirical Mode Decomposition(EMD)is a multi-scale adaptive decomposition method,and it is very beneficial for processing nonlinear and non-stationary signals.This method is regarded as an important innovation in signal processing and analysis,and has attracted extensive attention from researchers at home and abroad.Since EMD can efficiently represent the feature information of different scales of image data,it is beneficial to use EMD to improve the quality of image fusion.This paper doses some research on image fusion based on EMD and proposes a method of image fusion using two-dimensional unconstrained EMD based on Delaunay triangulation.It can not only preserve the detailed features of multi-source images well,but also enhance the fusion results through appropriate enhancement coefficients in the fusion process in order to obtain a higher quality fusion images.The specific work of this paper can be listed as follows:(1)A two-dimensional unconstrained EMD method based on Delaunay triangulation is proposed.The method firstly redefines the extremum points of a two-dimensional image,and then constructs an unconstrained optimization model using the Delaunay triangulation on the defined extremum points,to iteratively decompose the image,which can adaptively decompose the original image into several modal images with fine-coarse scales and a residue image.The experimental results show that the proposed method has stronger detail acquisition ability than the original two-dimensional unconstrained EMD method,and can better capture the different scale features of the original image.(2)An image fusion method based on unconstrained optimization EMD is presented.The method firstly decomposes the multi-source image by the two-dimensional unconstrained EMD method based on Delaunay triangulation,and then fuses the modal images at the same level and the residual image according to some specific fusion criteria.Furthermore,the fused image can be enhanced to obtain a higher quality fused image through appropriate enhancement coefficients in the fusion process.Experiment results on some typical multi-focus images,multi-spectral and panchromatic images,CT and MRI images demonstrate that the proposed method obtains good results.